DESCRIPTION (Applicant's abstract): This application seeks support for a team of statisticians, economists, clinicians, and mental health services researchers to collaborate on the development and application of discrete choice models for understanding treatment use and for causal inferences in experimental and naturalistic studies of mental illness. By studying how patients are matched with treatments in extant systems, researchers will gain greater insight into the determinants of quality of care. The Specific Aims will involve the 1) extension of likelihood-based methods to estimate treatment effectiveness at the levels actually received using experimental data from two influential clinical trials (Schulberg, Block, Madonia et al., Acrh Gen Psychiatry 1996;53:913-9 & Rosenheck, Neale, Arch Gen Psychiatry 1998;55:459-66) and to compare these estimates with those based on conventional approaches, such as intention-to-treat, adequate, and completer principles, 2) development of new models of discrete choice to explain variation in treatment use based on patient, provider, and insurance characteristics for privately insured and Medicaid beneficiaries, and 3) application of these discrete choice models to explain variation in adherence with treatment recommendations and in treatment effectiveness for depression and for schizophrenia across a diverse array of practice settings. An Advisory Board comprised of leaders in statistics, economics, and psychiatry will convene annually to validate methods and ensure integration of techniques into mental health services research. The methodological advances from this research will enable mental health researchers and policy makers to better characterize usual care and to expand the inferences drawn from clinical trials.